The argument is as follows.
1 For any two acts, one done by a human and one done by a non-human, if the one done by the non-human has worse outcomes, we should prefer the one done by the human.
2 A human killing one person is less bad than a non-human natural force killing two
Therefore, we should prefer a human killing one to a non-human natural force killing two
3 If A is preferable to B, we should prefer a world that has A and ~B to a world that has B but not A
Therefore, we should prefer a world that has a human killing one to a non-human natural force killing two
4 If a person kills one to prevent a non-human natural force from killing two, that would be a world that has a human killing one and doesn’t have a non-human natural force killing two
Therefore, we should prefer a world in which one kills one to prevent a non-human natural force from killing two to one in which they don’t
5 For any action N, if we should prefer a world in which N is taken to one in which N isn’t taken, then action N should be taken
Therefore, a person should kill one to prevent a non-human natural force from killing two.
Objections?
A New Objection To Rights
I take it that "rights" are supposed to be deontological constraints on actions, i.e. rights prohibit the ways that we are morally permitted to use other agents to make the world better. So deontologists would say that agents should respect rights even in cases where violating rights would make the world better (unless the benefits are extremely large, according to moderate deontologists).
I also take it that to say world A is "better" than world B is just to say that there is more reason to prefer A be instantiated rather than B be instantiated. This seems to be what we mean when we say one world is "better" than another.
If that's the correct meaning of "better", then premise (5) will be question-begging to the deontologist. We can translate the statement "For any action N, if we should prefer a world in which N is taken to one in which N isn’t taken, then action N should be taken" to the statement "For any action N, if the world in which N is taken is better than the world in which N isn't taken, then action N should be taken". But that's just an explicit negation of deontology.
I know that some writers argue that deontologists should maintain the view that we should want agents to respect rights, even in cases where violating rights would make the world better. These writers are presumably using a different definition of "better" than the one I mentioned. On this view of deontology, premise (1) would be question-begging since it's an explicit negation of the view.
I don't see the motivation for 1). Almost everyone has the intuition that deliberate acts are worse than natural events with the same outcome, eg murder is worse than being struck by lightning.